Molecular similarity concepts and search calculations

The introduction of molecular similarity analysis in the early 1990s has catalyzed the development of many small-molecule-based similarity methods to mine large compound databases for novel active molecules. These efforts have profoundly influenced the field of computer-aided drug discovery and substantially widened the spectrum of available ligand-based virtual screening approaches. However, the principles underlying the computational assessment of molecular similarity are much more multifaceted and complex than it might appear at first glance. Accordingly, intrinsic features of molecular similarity analysis and its relationship to other methods are often not well understood. This chapter discusses critical aspects of molecular similarity, an understanding of which is essential for the evaluation of method development in this field. Then it describes studies designed to enhance the performance of molecular fingerprint searching, which is one of the most intuitive and widely used similarity-based methods.